diff --git a/CMakeLists.txt b/CMakeLists.txt
index a8838cebc..9cd0b9183 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -12,8 +12,11 @@ IF (NOT CMAKE_BUILD_TYPE)
  SET(CMAKE_BUILD_TYPE "Release")
 ENDIF (NOT CMAKE_BUILD_TYPE) 
 
+SET(CMAKE_CXX_FLAGS_DEBUG  "-O0 -g -DDEBUG")
 
-
+SET(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -D_DEBUG=3")
+SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} -D_DEBUG=4")
+MESSAGE("CMAKE_CXX_FLAGS_DEBUG="+${CMAKE_CXX_FLAGS_DEBUG})
 OPTION(USE_DOUBLE_PRECISION "Use double precision"	OFF)
 OPTION(USE_GRAPHICAL_BENCHMARK "Use Graphical Benchmark" ON)
 OPTION(USE_MULTITHREADED_BENCHMARK "Use Multithreaded Benchmark" OFF)
diff --git a/Demos/ThreadingDemo/main.cpp b/Demos/ThreadingDemo/main.cpp
index 90ccb908a..e5168a76c 100644
--- a/Demos/ThreadingDemo/main.cpp
+++ b/Demos/ThreadingDemo/main.cpp
@@ -67,6 +67,8 @@ struct SampleThreadLocalStorage
 
 void	SampleThreadFunc(void* userPtr,void* lsMemory)
 {
+	printf("thread started\n");
+
 	SampleThreadLocalStorage* localStorage = (SampleThreadLocalStorage*) lsMemory;
 
 	SampleArgs* args = (SampleArgs*) userPtr;
@@ -138,10 +140,11 @@ int main(int argc,char** argv)
 		for (i=0;i<numThreads;i++)
 		{
 			threadSupport->waitForResponse(&arg0,&arg1);
+			printf("finished waiting for response: %d %d\n", arg0,arg1);
 		}
 	} else
 	{
-#if 0
+#if _WIN32
 		int numActiveThreads = numThreads;
 		while (numActiveThreads)
 		{
@@ -155,10 +158,15 @@ int main(int argc,char** argv)
 				printf("polling\n");
 			}
 		};
+#else
+	btAssert(0);
+	printf("non-blocking wait is not supported on this platform\n");
+	exit(0);
 #endif
 	}
 
-	threadSupport->stopSPU();
+printf("stopping threads\n");
+
 	delete threadSupport;
 	printf("Press ENTER to quit\n");
 	getchar();
diff --git a/src/BulletMultiThreaded/PosixThreadSupport.cpp b/src/BulletMultiThreaded/PosixThreadSupport.cpp
index b41bac6a8..5fc73c85e 100644
--- a/src/BulletMultiThreaded/PosixThreadSupport.cpp
+++ b/src/BulletMultiThreaded/PosixThreadSupport.cpp
@@ -48,7 +48,7 @@ PosixThreadSupport::~PosixThreadSupport()
 #endif
 
 // this semaphore will signal, if and how many threads are finished with their work
-static sem_t* mainSemaphore;
+static sem_t* mainSemaphore=0;
 
 static sem_t* createSem(const char* baseName)
 {
@@ -58,9 +58,10 @@ static sem_t* createSem(const char* baseName)
         char name[32];
         snprintf(name, 32, "/%s-%d-%4.4d", baseName, getpid(), semCount++); 
         sem_t* tempSem = sem_open(name, O_CREAT, 0600, 0);
+
         if (tempSem != reinterpret_cast<sem_t *>(SEM_FAILED))
         {
-        	//printf("Created \"%s\" Semaphore %x\n", name, tempSem);
+//        printf("Created \"%s\" Semaphore %p\n", name, tempSem);
         }
         else
 	{
@@ -199,7 +200,8 @@ void PosixThreadSupport::startThreads(ThreadConstructionInfo& threadConstruction
 	m_activeSpuStatus.resize(threadConstructionInfo.m_numThreads);
         
 	mainSemaphore = createSem("main");                
-        
+	//checkPThreadFunction(sem_wait(mainSemaphore));
+   
 	for (int i=0;i < threadConstructionInfo.m_numThreads;i++)
 	{
 		printf("starting thread %d\n",i);
@@ -233,15 +235,23 @@ void PosixThreadSupport::startSPU()
 ///tell the task scheduler we are done with the SPU tasks
 void PosixThreadSupport::stopSPU()
 {
-	for(size_t t=0; t < size_t(m_activeSpuStatus.size()); ++t) {
+	for(size_t t=0; t < size_t(m_activeSpuStatus.size()); ++t) 
+	{
             btSpuStatus&	spuStatus = m_activeSpuStatus[t];
             printf("%s: Thread %i used: %ld\n", __FUNCTION__, int(t), spuStatus.threadUsed);
-        
-            destroySem(spuStatus.startSemaphore);
-            checkPThreadFunction(pthread_cancel(spuStatus.thread));
-        }
-        destroySem(mainSemaphore);
 
+	spuStatus.m_userPtr = 0;       
+ 	checkPThreadFunction(sem_post(spuStatus.startSemaphore));
+	checkPThreadFunction(sem_wait(mainSemaphore));
+
+	printf("destroy semaphore\n"); 
+            destroySem(spuStatus.startSemaphore);
+            printf("semaphore destroyed\n");
+		checkPThreadFunction(pthread_cancel(spuStatus.thread));
+        }
+	printf("destroy main semaphore\n");
+        destroySem(mainSemaphore);
+	printf("main semaphore destroyed\n");
 	m_activeSpuStatus.clear();
 }